French people and AI in healthcare: curiosity, caution and fault lines

At a time when diagnostics, research and hospital administration are beginning their technological shift, how do the French welcome this transformation of one of the most sensitive sectors of their daily lives? Do they have confidence in the capabilities of AI? What role are they prepared to give it in the care process? Do they believe that AI will one day supplant doctors' skills? Have they themselves already put into practice advice provided by ChatGPT, Claude or Google Gemini?

Between curiosity and caution, emerging uses and deep-seated mistrust, the study whose results we are unveiling paints a picture of divided opinion. Generational divides, differences in apprehension between men and women, demands for transparency - perceptions vary, sometimes sharply, but converge on the same widely shared conviction: for the majority, AI in healthcare should only be a tool at the service of the doctor, never a substitute for the human relationship at the heart of care.

A definite but still vague reputation

More than 6 out of 10 people (64%) claim to have heard of artificial intelligence applied to healthcare. In detail, 10% claim to have in-depth knowledge, while 54% have a more vague perception. Conversely, 36% have never heard of it.

A closer look at the results reveals a clear gender divide: 70% of men say they have heard of AI in healthcare, compared to 58% of women. This 12-point gap suggests a persistent inequality in access to scientific and technological information, or even less interest or exposure to topics related to medical innovation among women.

A credibility that remains fragile...

When asked how much confidence they have in the answers given to health-related questions by generative AI such as ChatGPT, Claude or Google Gemini, the French are divided: 43% give these tools some credibility, while 45% are distrustful. Those who are completely confident are four times less numerous than those who are totally distrustful (4% vs. 16%).

The breakdown by age group reveals a very marked trend: probably because they master these tools better, more than half (56%) of 18-24 year-olds say they trust these technologies, compared with only a third (33%) of those aged 65 and over. Men are also more convinced than women (51% vs. 36%), figures that confirm the gaps already observed in levels of information.

... but a high potential for take-up

For the time being, the personal use of generative AI in healthcare remains in the minority: 34% of respondents say they have already used it. But 38% of non-users say they are ready to do so in the future, while only 28% reject the prospect entirely. Proof that, despite their relatively recent emergence, these tools have already gained considerable ground and will most likely continue to do so in the near future.anté, the French are divided: 43% give these tools some credibility, while 45% are wary. Those who are completely confident are four times less numerous than those who are totally distrustful (4% vs. 16%).

This is all the more likely as the younger generations have taken to it en masse: 68% of 18-24 year-olds say they have used generative AI to ask a health question, compared with just 10% of the over-65s.

Advice followed... after medical validation

Among those who have ever asked an artificial intelligence questions on health-related topics, six out of ten (60%) say they followed the recommendations they received. Most (43%) say they did so after validation by a professional, but 17% say they applied them without consulting a doctor. These figures illustrate a dual reality: on the one hand, AI is becoming a real source of health information and advice, but on the other, validation by a practitioner remains an obligatory step for the majority of users. What's more, 34% of those seeking answers did not act on the advice given to them by AI.

Expectations focused on diagnostics and research

When asked about the most useful contributions of artificial intelligence to healthcare, the French clearly place two areas at the top of the list: disease diagnosis (48%) and the search for new treatments (47%). These results illustrate the high expectations for AI's ability to improve the accuracy and effectiveness of healthcare. Next come

more functional, but equally useful uses, such as administrative automation (37%) or epidemic surveillance (31%). Only 13% of respondents consider that artificial intelligence brings no added value to the healthcare sector, with women (16%) and the over-65s (17%) being the most numerous in this case.

Support for a divisive decision

The idea that their doctor could rely on artificial intelligence to refine a diagnosis or propose a treatment also polarizes respondents. 49% expressed a feeling of confidence (of which only 6% were very reassured), while 34% were worried and 17% had no opinion.

Women are clearly more worried than men (41% vs. 27%), a gap that reinforces the idea of a more distant or distrustful relationship on their part towards automated technologies in the medical field.

A massive need for transparency

AI cannot be integrated into the care pathway without transparency about its use. This is confirmed almost unanimously by those polled: 80% feel it is important to be informed upstream when AI is involved in their care. Among them, 41% even want to know precisely when and how it will be used. This expectation is shared in similar proportions across all generations and genders. Ethical guidance and clarity of practice will therefore be decisive in promoting the acceptability of artificial intelligence in healthcare in the years to come.

AI superiority: yes, but limited to certain fields

While only a small proportion (12%) of French people believe that artificial intelligence will become more reliable than doctors, more than half (53%) recognize that it could surpass humans, but only in certain technical areas. Here again, the idea of assistance taking advantage of AI's capabilities outweighs that of replacement. However, 28% of those questioned are convinced that human expertise will always remain superior, while 10% reject the hypothesis of AI supremacy, on the grounds that it presents too great a risk of error.

Dehumanization of the care relationship at the heart of fears

When asked to identify their main fear linked to the use of artificial intelligence in healthcare, the French cited the dehumanization of the link with the doctor (34%), followed by the risk of errors (28%) and the lack of human control (24%). Concerns about confidentiality are a distant second (7%), as is a lack of concern (also 7%). This hierarchy of responses confirms that the human link remains paramount, even when the technological benefits are recognized.

AI less forgiving of error

For 53% of respondents, no misdiagnosis is acceptable, whether it comes from a doctor or an AI.  Women (59%) are more intransigent than men (47%). Nearly one in five (18%), on the other hand, believe that such errors are acceptable, whatever their origin.

On the other hand, 20% are more forgiving of human error, compared with 9% who would be more tolerant of AI error. Young people (aged 18-24) are more inclined than average to accept an AI diagnostic error (17%, versus 9%).

Operating without a doctor: men more confident than women

30% of French people say they are ready to trust an AI to manage a surgical operation on its own, including 8% without hesitation and 22% if the operation is not very complex. On the other hand, 40% want a doctor to be present, and 22% are formally opposed.

Twice as many men as women accept this prospect (40% versus 21%), and young adults are slightly more open to it than their elders (35% among 18-24 year-olds and 26% among the over 65s). Overall, however, unmanned surgery is still widely rejected.

Survey conducted by FLASHS for Galeon from March 11 to 14, 2025 by self-administered online questionnaire among a panel of 2,003 French men and women aged 18 and over, representative of the French population.